Site Home   Archive Home   FAQ Home   How to search the Archive   How to Navigate the Archive   
Compare FPGA features and resources   

Threads starting:
1994JulAugSepOctNovDec1994
1995JanFebMarAprMayJunJulAugSepOctNovDec1995
1996JanFebMarAprMayJunJulAugSepOctNovDec1996
1997JanFebMarAprMayJunJulAugSepOctNovDec1997
1998JanFebMarAprMayJunJulAugSepOctNovDec1998
1999JanFebMarAprMayJunJulAugSepOctNovDec1999
2000JanFebMarAprMayJunJulAugSepOctNovDec2000
2001JanFebMarAprMayJunJulAugSepOctNovDec2001
2002JanFebMarAprMayJunJulAugSepOctNovDec2002
2003JanFebMarAprMayJunJulAugSepOctNovDec2003
2004JanFebMarAprMayJunJulAugSepOctNovDec2004
2005JanFebMarAprMayJunJulAugSepOctNovDec2005
2006JanFebMarAprMayJunJulAugSepOctNovDec2006
2007JanFebMarAprMayJunJulAugSepOctNovDec2007
2008JanFebMarAprMayJunJulAugSepOctNovDec2008
2009JanFebMarAprMayJunJulAugSepOctNovDec2009
2010JanFebMarAprMayJunJulAugSepOctNovDec2010
2011JanFebMarAprMayJunJulAugSepOctNovDec2011
2012JanFebMarAprMayJunJulAugSepOctNovDec2012
2013JanFebMarAprMayJunJulAugSepOctNovDec2013
2014JanFebMarAprMayJunJulAugSepOctNovDec2014
2015JanFebMarAprMayJunJulAugSepOctNovDec2015
2016JanFebMarAprMayJunJulAugSepOctNovDec2016
2017JanFebMarAprMayJunJulAugSepOctNovDec2017
2018JanFebMarAprMayJunJulAugSepOctNovDec2018
2019JanFebMarAprMayJunJulAugSepOctNovDec2019
2020JanFebMarAprMay2020

Authors:A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Custom Search

Messages from 5025

Article: 5025
Subject: ASIC/FPGA Synthesis for LINUX... It's HERE!
From: "root@speedy.login.qc.ca" <root@speedy.login.qc.ca>
Date: 14 Jan 1997 04:39:16 GMT
Links: << >>  << T >>  << A >>
This is a multi-part message in MIME format.

--------------7B26476967E668C57805B
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

See the attachment for a comparison of Win95 & Linux 2.0.27 on synthesis
of the same VHDL circuits.

--------------7B26476967E668C57805B
Content-Type: text/plain; charset=us-ascii; name="linux_leo.txt"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline; filename="linux_leo.txt"

A modest comparison of Windows-95 against Linux 2.0.27

by John McCluskey,   J-Squared Technologies.              January 8, 1997
Email:    J.McCluskey@ieee.org

This experiment started on Monday night (Jan 6), when after a phone call to
Exemplar Logic in California, they agreed to place the Linux version of
their Leonardo synthesis tool on their ftp site so I could grab it to test
it out.  I spent the rest of the night (until 1 am) dialing and redialing my
ISP's modem pool, trying to latch onto a good modem that would hold the
modem carrier for more than a few minutes.  After a lot of frustration, I
gave up and finished grabbing the rest of the gzipped tar file in the
morning.  Finally, about 10 AM, I had succeeded in grabbing the 15 megabyte
file, gotten a license code from Exemplar, and I was ready to Rock & Roll!.

Leonardo is a VHDL/Verilog synthesis tool that compiles a source language
like VHDL or Verilog into a gate level netlist suitable for input into a
place & route tool that produces a final design ready for use.  The place &
route tool might be for a CPLD (Complex Programmable Logic Device), or an
FPGA (Field Programmable Gate Array), or for a genuine ASIC (Application
Specific Integrated Circuit).  Leonardo does them all, and currently runs on
HPUX, Solaris 2.5, SunOS 4.1.3, Windows 95/ WinNT, and now, LINUX!  This is
a genuine industrial strength CAD tool.  The graphical interface is TCL/Tk
based on ALL platforms.  The user interface is extensible, since you can add
buttons or modify them by playing with the TCL scripts which generate the
GUI.  Indeed, the whole synthesis process is driven by TCL scripts.  I
generated a base TCL script for each test using the GUI, and then used the
script in batch mode to get my timing tests for Linux, SunOS 4.1.3, and
Win95.  Regrettably, I didn't have time to test Windows NT4.0.  This Linux
version of Leonardo was compiled on a Debian Linux distribution using GCC
2.72 or GCC 2.73.  Note that Pentium specific optimization is NOT included
in these versions of GCC.  (I think.)

The first test machine was a Sparc20 clone from Axil, dating from June 1995,
which has 128 Mb of RAM, and a 90 MHz Ross Hypersparc processor module with
1 Mb of cache.  The disk subsystem is SCSI based, but this hardly matters,
since with 128 Mb of memory, it never swapped into the 256 Mb swap space.
The CPU is pretty slow compared to a Pentium.  The sparcstation is useful as
a reference point since I was able to get a memory usage estimate from
'time', which gave me the working set size (wss).  The second test machine
is a Pentium 120 PC, with an ASUS Triton 2 motherboard with 512K cache and
dual NEC 2Gb IDE disk drives.  It was first tested with 32 Mb of 60 ns DRAM,
and then upgraded to 64 Mb.  For some of the tests, memory makes a big
difference, since logic synthesis can be a CPU and memory hog of the worst
sort.  On the Pentium PC, two operating systems were tested: Windows 95 vs
Linux with kernel 2.0.27 and a Slackware 3.1 (Slackware96) distribution.

For testing the compiler, I dug up some VHDL oinkers of extensible
proportions.  In a VHDL compiler, compilation proceeds in 3 phases: Parsing,
Elaboration, and Optimization.  Parsing is just about the same as a C++
compiler, but in Elaboration, the circuit design specified by the behavorial
VHDL code is translated into large parallel data structures that describe
the boolean behavior of the code.  In elaboration, all loops are unrolled
and constants are propagated as far as possible.  Memory usage often hits
the first peak during elaboration.  The next phase, optimization, has the
job of compressing the vast parallel data structures into actual gates that
describe the same behavior.  The objective in optimization is often to
minimize the circuit area (minimize gate count), but is often subject to a
constraint that the circuit delay be less than some specified minimum
(constrained synthesis).  During optimization, memory usage can often spike
dramatically higher, depending on the algorithm being used.  The optimizer
has 8 or 9 different algorithms at it's disposal, and standard procedure is
to run them all, and then save the best result.  This can burn a lot of CPU
time.

Test #1 (Serpent) is a simple shift register for an ORCA 2C FPGA, but with a
twist.  The length of the shift register is set by a parameter, and the
flip-flop elements are directly instantiated with a loop in the code.  Each
flip-flop element has an attribute attached that directs the place & route
tool to place the flip-flop in a specified location on the FPGA die.  The
code places the flip-flops so that the shift register is done with a
serpentine layout in a block of cells.  Technically, this is known as
'floorplanning' and is usually done with a GUI tool.  For reason's not yet
clear, this bit of code is an enormous memory hog, and uses memory in
proportion to the number of elements instantiated.  The number of flip flops
is set to 640, which are placed in 160 separate locations This test does
elaboration only, since there is no point in optimization.  The synthesis
script just loads the gate library, reads in and elaborates the code, then
just writes out the resulting netlist.

When compiled on the Sparcstation, this took 2 minutes 27 seconds of wall
time, and hit a working set size peak (memory usage) of 62 Megabytes.  Under
Linux 2.0.27, with 32 Mb of RAM and a 96 Mb swap space, it thrashed badly
but still produced a result after 25 minutes and 30 seconds.  Windows 95
died with an "Out of Memory" message after thrashing for 18 minutes and 54
seconds.  Another 32 Mb of RAM was added to the Pentium and this was
repeated.  Linux improved its time to 13 minutes even, while Windows 95
managed to get a result after 17 minutes, 20 seconds.  Linux has a pretty
good advantage here, since the compiler can run without bothering to load X
windows, while under Win95, the compiler won't run unless the graphical
interface is up.  The compiler can't run in DOS mode.  For the Windows 95
tests, a DOS window was opened and the compiler batch file was run in that
window.  Nothing else was running.

Test #2 (CRC) is a CRC (Cyclic Redundancy Code) calculation on a parallel
stream of data words.  CRC calculation is usually done on a serial data
stream, but what if the data stream is a presented as a sequence of parallel
words?  This design, targeted for Xilinx 4000E FPGA's, feeds the incoming
word to the address of lookup table which contains the bits needed to XOR
with the current CRC data word stored in a data register.  This design is
also extensible, since both the CRC polynomial and the input data word width
are specified as parameters.  I used the CRC-CCIT polynomial and a 12 bit
input data word.  This generates a 4096x16 table in elaboration, which is
then collapsed down into a small number of gates in optimization.  This
works because the table is sparse and has a low entropy.  This design has
significant amounts of work to do in both elaboration and in optimization.

On the Sparc, this compiled in 6 minutes 11 seconds, using 25 Mb of RAM.
Linux on the Pentium 120 with 32 Mb RAM compiled in 4 minutes 37 seconds.
Windows 95 trailed at 7 minutes 18 seconds.

Test #3 (2901) is a design with not much elaboration, and mostly
optimization work to do.  It is the venerable 2901 bit slice design,
snatched straight from the VHDL.ORG website.  This is basically a 4 bit
processor slice that does addition, subtractions and other binary operations
on a 4 bit data path.  I didn't write the code for this, and I would
certainly write it differently if I had the time for it.  It does give the
compiler something to chew on when I compiled it for Altera Flex10K FPGA's.

On the Sparc, this took 17 minutes 37 seconds in 18.5 Mb of RAM, while Linux
chewed this up in 8 minutes, 53 seconds.  Again, Windows 95 took second
place in 12 minutes and 37 seconds.

Table #1:   The following table has the run times shown in seconds.

  Test        Sparc RAM Usage   Sparc20     Windows 95    Linux 2.0.27
----------------------------------------------------------------------
#1 (Sparc-20)     62 Mb         147 sec
#1 PC 32 Mb RAM                                died          1530 sec
#1 PC 64 Mb RAM                              1040 sec         780 sec

#2 CRC            25 Mb         371 sec      438 sec          277 sec

#3 2901           18.5 Mb      1057 sec      757 sec          533 sec

If you take the Windows 95 run-times and sum together, and then divide by
the sum of the Linux run-times, you get 1.405.  What does this mean?  Linux
is 40% faster than Windows 95.  Or to put it another way, you can turn up
your effective clock speed by 40% when you run Linux.  I like it.  Does this
still hold true for Windows NT4.0?  Good question.  Stay tuned.

Conclusions: Linux looks very good as an alternative to Windows 95, but for
tough problems, you can never have too much RAM or Swap space!  A Pentium
CPU easily beats the older Sparc CPU's and a Pentium Pro CPU should be able
to trounce all but the priciest of workstations.  The price performance of a
Pentium Pro is probably unbeatable.

I called up Exemplar Logic and queried them further on their policies of
making the Linux binary available to their customers...  It turns out that
anyone who buys a floating license can run on whatever platform they please,
whether it is an HP, Sparc, Win NT, Win95, or Linux, or all of the above.
The linux version can grab a license token from the server just like the HP
or Sparc version.  (I've gotta try this!) Indeed, even the Win95 version can
grab a license token if it is configured properly.  Linux can even be used
for the license server, since the FlexLM binaries work just fine.  The only
platform not supported as a license server is Win95!  :-)

Summary:  Win95 bytes,  Linux cruises to victory...

Next Up:  The crusade for a Linux Place & Route tool begins.


To answer the final obvious question:

             Where can I get a linux binary of Leonardo?

Answer:  You can't.   At least for a week while they complete the final
         rebuilds of Leonardo 4.0.3 on all the platforms.  The new version
         should be on the (protected) FTP site by the end of January 97.
         I'll post a notice when it's ready.   If you have access to a Leonardo
         license server, contact me and I'll put you on a mailing list to
         get early access.

John McCluskey
J-Squared Technologies

J.McCluskey@ieee.org

--------------7B26476967E668C57805B--

Article: 5026
Subject: Re: Any PEEL22CV10A replacements with more capacity?
From: Geoffrey Bostock <geoff.bostock@zetnet.co.uk>
Date: Tue, 14 Jan 1997 11:03:23 GMT
Links: << >>  << T >>  << A >>
In message <32DA860A.2247@aspect.com>
        "Thomas C. Jones" <tjones@aspect.com> writes: 

> Does anyone know of a pin-compatible replacement to the PEEL22CV10A 
> offering greater routing capacity and resources?

> -Tom
> tjones@aspect.com

Why not try the Philips PLC42VA12 or Lattice GAL6001/2.   These are 
both FPLA architecture and have less P-terms in total (64 for each) 
but they are much more efficient for state machines and the 6001/2 
have buried registers.

Geoff Bostock

(Author Programmable Logic Handbook and FPGAs & Programmable LSI - 
PLD/FPGA Design Consultant)



Article: 5027
Subject: Re: Any PEEL22CV10A replacements with more capacity?
From: Steve Wiseman <steve@sj.co.uk>
Date: Tue, 14 Jan 1997 12:32:20 +0000
Links: << >>  << T >>  << A >>
Thomas C. Jones wrote:
> 
> Does anyone know of a pin-compatible replacement to the PEEL22CV10A
> offering greater routing capacity and resources?

ICT do the PA7024I PEEL array, which, quoting from my '94 databook...
"Most powerful 24-pin PLD available"  (In 1994, believable)
20 I/Os, 2 Inputs / clocks, 40 registers / latches
speeds to 17nS (In 1994. Now, who knows)
blah blah blah (seems to offer all the funky bits of a PEEL22CV10A+, so
looks like a fine device. )
Also, if you've been using ICTPLACE to code your designs, it's supported
in there, so you get the simulator and the decent design tools. 
  (Side note, that simulator is what gets me to design in ICT PLDs
rather than generics. Well worth the effort, ICT, Thanks. (In fact the
whole development environment for ICT is pleasant)

Steve
Article: 5028
Subject: Re: Great Xilinx FPGA Kits & prices
From: timolmst@cyberramp.net
Date: Tue, 14 Jan 1997 14:52:27 GMT
Links: << >>  << T >>  << A >>
"James Fakas" <jjfakas@erols.com> wrote:

>These are extremely good deals. They include a PC ISA FPGA test board board
>along with the FOUNDATION software from XILINX!

>Great deals!!!

>Here are other kit prices:

>APS-X84-FB X84 board with the Foundation Base Software....$650.00

>APS-X84-FBV same with VHDL............$1295.00

>APS-X84-FS X84 board with Foundation Standard Software....$1300.00

>APS-X84-FSV same with VHDL..............$2500.00

>The contacts are at

I believe that the Xilinx Foundation software is actualy an evaluation
copy, and it turns into a pumpkin after 250 uese.

Article: 5029
Subject: Re: DES Keysearch by FPGA: $10,000 prize
From: Alain RAYNAUD <Alain_Raynaud@mentorg.com>
Date: Tue, 14 Jan 1997 15:58:07 +0100
Links: << >>  << T >>  << A >>
Peter Trei wrote:
> 
> In 1993, Michael J. Wiener of Bell-Northern Research, published a
> paper 'Efficient DES Keysearch', in which he gave a detailed
> description of such a keysearch engine, based on custom chips, which
> could search the entire keyspace in about 7 hours.

Is a gate-level netlist or a VHDL or Verilog description of this design
available somewhere ?

I'd like to try it, but don't really have the time to do a capture from
the article. Since it is public domain, surely somebody has it
(including the author of the article I guess).

Alain.
-- 
-----------------------------------------------------------------------
 Alain RAYNAUD                                            META SYSTEMS 
                                                       Batiment Hermes 
                                                     4, rue Rene Razel 
 Tel: (33) 01 69 35 10 16                        91400 Saclay - FRANCE
 E-Mail: Alain_Raynaud@mentorg.com            Fax: (33) 01 69 35 10 10 
-----------------------------------------------------------------------
Article: 5030
Subject: ANNOUNCE: New Model and Tip of the Month
From: rob@doulos.co.uk (Rob Hurley)
Date: Tue, 14 Jan 1997 17:39:17 GMT
Links: << >>  << T >>  << A >>
ANNOUNCE: NEW MODEL AND TIP OF THE MONTH

This month's model is:
	
		Simple RAM model in Verilog

This month's tip is:
	
		Re-usable Functions

You can find both at 'THE WINNING EDGE'   http://www.doulos.co.uk 

You can also access previous Models and Tips of the Month from the 
same site. During 1996 these included:
		A to D converter
		Design for Debug
		How to avoid synthesising unwanted latches 
		FIR filter
				.... and many more

During 1997 Doulos will continue to develop THE WINNING EDGE to 
include more VHDL and Verilog tips, tricks, tutorials and models.

Our aim is to make THE WINNING EDGE one of the most useful High 
Level Design sites out there.

____________________________________________________________________

Also *** NEW *** for this month are:

	Free CBT software for Verilog

	Additions to our Hardware Engineer’s Guide to VHDL series
_____________________________________________________________________



DOULOS
Church Hatch				Tel: +44 1425 471 223
22 Market Place				Fax: +44 1425 471 573
Ringwood BH24 1AW			Email: webmaster@doulos.co.uk
UK

_____________________________________________________________________






Article: 5031
Subject: Design Automation Conference Scholarships
From: cohoon@mamba.cs.Virginia.EDU (Jim Cohoon)
Date: Tue, 14 Jan 1997 18:45:59 GMT
Links: << >>  << T >>  << A >>
The following information is also available at
        http://www.dac.com/dac/grant.html



DAC DESIGN AUTOMATION GRADUATE SCHOLARSHIPS

The 34th Design Automation Conference is sponsoring up to four $12,000
scholarships to support graduate research and study in Design Automation,
with emphasis in electronic and computer design and test automation.
These scholarships are awarded directly to a university for the Faculty
Investigator to expend in accordance with the proposal.

The university receiving a scholarship is free to use this money in
direct support of one or more of its Design Automation graduate students
named in the proposal, in the manner outlined in its proposal, except
that the scholarship funds shall not be used to support indirect costs
or overhead.

The scholarship funds must only be used to support the student(s) and
project(s) named in the winning proposals. Because the award decisions
are made based on the scholastic achievements of the named student(s)
and the qualifications of the named project(s), no substitutions may
be made. The scholarship recipient must forfeit the award if the conditions
of the research proposal change. The Faculty Investigator of the project
receiving a scholarship award will be expected to submit a brief report
of the year's activities supported by the scholarship for publication
in the newsletters of the sponsoring organizations. This report should
be submitted at the conclusion of the year supported. Applications for
the scholarship, either new or renewal, should be submitted by the Faculty
Investigator or Department Chairperson and must include:

  1. A brief (2 pages maximum) biography of the student(s) proposed for
     the scholarship;

  2. Transcripts of the graduate and undergraduate records of each of the
     student(s) proposed for the scholarship;

  3. A paragraph for each student proposed outlining his or her goals and
     objectives;

  4. A brief (5 pages maximum) proposal concerning the research to be
     conducted and the way in which the scholarship is to be expended;

  5. A brief statement (1 page maximum) of the impact of the scholarship
     on the Design Automation Program at her or his institution;

  6. A listing of all current and pending support for the student(s) and
     project(s): who the support is from, the duration of the support,
     and the dollar amount of the support.

  7. A indication of any previous Design Automation Scholarships awarded
     to the Faculty Investigator. 

Scholarships will be awarded based on the following criteria:

  *  Academic credentials of the student(s);

  *  Quality and applicability of the proposed research;

  *  Impact of the award on the DA program at the institution. Preference
     will be given to institutions which are trying to establish new DA
     research programs. 

  *  Need. Preference will be given to student(s) and project(s) that can
     demonstrate financial need. 

Scholarships can be awarded in support of new projects or for one renewal
of a DAC grant from the previous year. An application from the same student,
or faculty sponsor of a winner from the previous year is considered a
renewal application.

Original and two complete copies of all application materials must be
received by Friday, January 31, 1997 to

     James P. Cohoon 
     University of Virginia 
     Department of Computer Science 
     Olsson Hall 
     Charlottesville, VA 22903 

For further information contact James Cohoon at + (804) 982-2210 or by
email at cohoon@virginia.edu. Notification of the awards will be made by
mid-April, 1997.
Article: 5032
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: Brad Taylor <blt@emf.net>
Date: Tue, 14 Jan 1997 13:53:08 -0800
Links: << >>  << T >>  << A >>
I'm fairly ignorant of radiation issues and I have a few questions for
anyone who might care to respond.  I'm getting a little lost with all
the qualitative responses I've seen posted to this thread and need a
little reality check (ie numbers).

- Approximately what failure rate would a non radhard FPGA die
(100mm^2)  experience at sea level?
- At 50,000 feet?
- In space?
- What happens to the above rates if the device is radhard?
- Why don't SRAM based FPGA manufacturers publish the expected rates?
- What is the effect of shielding?

- I've read that some cosmic rays contain as much energy as a well
thrown rock. Do radhard devices tolerate such hits?

If the expected rate is so low as to occur only infrequently, then some
simple means might be used to insure system integrity by simply
detecting the possibility of an error.

- Is it possible to monitor the die to detect possible hits by looking
at die leakage or other measurable parameters.  If so, could one detect
a possible hit and alert the system to reboot the FPGA?

- Could the FPGA architecture be altered to allow parity be computed
continously for config data to determine if it has been hit?

Just wondering
-
Brad
  the system.
Article: 5033
Subject: Re: ASIC/FPGA Synthesis for LINUX... It's HERE!
From: Uwe Bonnes <bon@hertz.ikp.physik.th-darmstadt.de>
Date: 14 Jan 1997 22:39:10 GMT
Links: << >>  << T >>  << A >>
In comp.os.linux.development.apps "root@speedy.login.qc.ca" <root@speedy.login.qc.ca> wrote:
: This is a multi-part message in MIME format.

: --------------7B26476967E668C57805B
: Content-Type: text/plain; charset=us-ascii
: Content-Transfer-Encoding: 7bit

: See the attachment for a comparison of Win95 & Linux 2.0.27 on synthesis
: of the same VHDL circuits.

: --------------7B26476967E668C57805B
: Content-Type: text/plain; charset=us-ascii; name="linux_leo.txt"
: Content-Transfer-Encoding: 7bit
: Content-Disposition: inline; filename="linux_leo.txt"

: A modest comparison of Windows-95 against Linux 2.0.27

: by John McCluskey,   J-Squared Technologies.              January 8, 1997
: Email:    J.McCluskey@ieee.org

: This experiment started on Monday night (Jan 6), when after a phone call to
: Exemplar Logic in California, they agreed to place the Linux version of
: their Leonardo synthesis tool on their ftp site so I could grab it to test
: it out.  I spent the rest of the night (until 1 am) dialing and redialing my
: ISP's modem pool, trying to latch onto a good modem that would hold the
: modem carrier for more than a few minutes.  After a lot of frustration, I
: gave up and finished grabbing the rest of the gzipped tar file in the
: morning.  Finally, about 10 AM, I had succeeded in grabbing the 15 megabyte
: ...

: Summary:  Win95 bytes,  Linux cruises to victory...

: Next Up:  The crusade for a Linux Place & Route tool begins.


: To answer the final obvious question:

:              Where can I get a linux binary of Leonardo?

: Answer:  You can't.   At least for a week while they complete the final
:          rebuilds of Leonardo 4.0.3 on all the platforms.  The new version
:          should be on the (protected) FTP site by the end of January 97.
:          I'll post a notice when it's ready.   If you have access to a
:	   Leonardo 
:          license server, contact me and I'll put you on a mailing list to
:          get early access.

Has Leonardo some reference page? Some mail adress?

Thanks
-- 
Uwe Bonnes                bon@elektron.ikp.physik.th-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. 06151 162516 -------- Fax. 06151 164321 ----------
Article: 5034
Subject: Re: Efficient DES Keysearch
From: trei@ziplink.net (Peter Trei)
Date: 14 Jan 1997 18:00:34 -0500
Links: << >>  << T >>  << A >>
In article <32DAF3D3.5EFA@dimensional.com> hamilton@dimensional.com writes:
>> Peter Trei wrote:
>> >
>> > In 1993, Michael J. Wiener of Bell-Northern Research, published a
>> > paper 'Efficient DES Keysearch', in which he gave a detailed
>           ^^^^^^^^^^^^^^^^^^^^^^^
>
>Is this paper available on the net somewhere ??

http://www.aist-nara.ac.jp/Security/doc/ is one location. It's in 
postscript - you'll need the graphics.

Peter Trei
ptrei@acm.org

Article: 5035
Subject: Newsgroups for ASIC design and tools?
From: brian@colloquium.co.uk (Brian Drummond)
Date: Wed, 15 Jan 1997 00:09:31 GMT
Links: << >>  << T >>  << A >>
Are there any newsgroups for the discussion of ASICs and custom silicon,
and ASIC vendors, VHDL libraries and design tools?
Please copy any replies to me in email.

Thanks in advance,

- Brian
Article: 5036
Subject: Re: Great Xilinx FPGA Kits & prices
From: "Richard Schwarz" <aaps@erols.com>
Date: 15 Jan 1997 04:27:08 GMT
Links: << >>  << T >>  << A >>
Not true!!!! These prices are for the NO RUN LIMITS software

Richard Schwarz
Pres.
APS

timolmst@cyberramp.net wrote in article
<5bg63k$s7d$2@newshost.cyberramp.net>...
> "James Fakas" <jjfakas@erols.com> wrote:
> 
> >These are extremely good deals. They include a PC ISA FPGA test board
board
> >along with the FOUNDATION software from XILINX!
> 
> >Great deals!!!
> 
> >Here are other kit prices:
> 
> >APS-X84-FB X84 board with the Foundation Base Software....$650.00
> 
> >APS-X84-FBV same with VHDL............$1295.00
> 
> >APS-X84-FS X84 board with Foundation Standard Software....$1300.00
> 
> >APS-X84-FSV same with VHDL..............$2500.00
> 
> >The contacts are at
> 
> I believe that the Xilinx Foundation software is actualy an evaluation
> copy, and it turns into a pumpkin after 250 uese.
> 
> 
Article: 5037
Subject: Re: Great Xilinx FPGA Kits & prices
From: timolmst@cyberramp.net
Date: Wed, 15 Jan 1997 13:44:34 GMT
Links: << >>  << T >>  << A >>
"Richard Schwarz" <aaps@erols.com> wrote:

>Not true!!!! These prices are for the NO RUN LIMITS software

>Richard Schwarz
>Pres.
>APS

If you say so! If so, then these are great prices.


Article: 5038
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: David Erstad <erstad@ssec.honeywell.com>
Date: Wed, 15 Jan 1997 08:33:28 -0600
Links: << >>  << T >>  << A >>
I think most people and designing and flying space systems would be very 
uncomfortable with the concept that a part would latch up and it would
be "OK".

Henry Spencer wrote:
> 
> In article <01bc0161$d44246a0$6e0db780@Rich> "Rich K." <rich.katz@gsfc.nasa.gov> writes:
> >...[DRAMs in space]  But making this work
> >(without too much trouble) requires two things: the device does not latchup
> >up from single particles (SEL or single event latchup) and a faborable
> >internal organization of the memory.  For the first, fortunately, there are
> >a number of available devices which are sufficiently immune to SEL...
> 
> Actually, if memory serves, the first requirement isn't quite as stringent
> as you might think:  it is typically possible to power-cycle DRAM briefly
> to break a latchup without losing the memory contents.  (Particularly if
> you are willing to do some testing and selection of parts, commercial DRAMs
> generally exceed their specs for data retention and refresh interval by a
> fairly wide margin.)  Admittedly, you don't want to do it *too* often.
> --
> "We don't care.  We don't have to.  You'll buy     |       Henry Spencer
> whatever we ship, so why bother?  We're Microsoft."|   henry@zoo.toronto.edu

-- 
Dave Erstad
erstad@ssec.honeywell.com
The views above are mine and are not necessarily Honeywell's.
Article: 5039
Subject: Special Issue of TVLSI, Prereg deadline for FPGA'97
From: hauck@eecs.nwu.edu (Scott A. Hauck)
Date: Wed, 15 Jan 1997 11:29:04 -0600
Links: << >>  << T >>  << A >>
Please find below the call for papers for the special issue of IEEE
Transactions on VLSI Systems on Recent Advances in FPGA Technology.

Also, please note that the preregistration deadline for the ACM/SIGDA
Symposium on Field-Programmable Gate Arrays is January 22nd, one week from
today.  For more details, please see:

htttp://www.ece.nwu.edu/~hauck/fpga97

Scott Hauck
hauck@ece.nwu.edu
==================================================================

                        CALL FOR PAPERS

         SPECIAL ISSUE ON RECENT ADVANCES IN FPGA TECHNOLOGY

            IEEE TRANSACTIONS ON  VLSI SYSTEMS


Field-Programmable Gate Arrays (FPGAs) have become  an  important
technology  for  the  implementation of VLSI circuits and systems
due to the steady increase in FPGA density and speed. Aside  from
main-stream  use in reasonably high volume production, the field-
programmability and re-programmability of FPGAs enable many novel
and  promising  applications, including rapid system prototyping,
reconfigurable system designs, rapid system prototyping,  circuit
emulation, and reconfigurable custom computing.  This special is-
sue will be devoted to recent advances in all  areas  related  to
the FPGA technology.

Topics of interest include, but not limited to:

o Advances in FPGA architectures, including design  of  programm-
  able logic blocks, programmable interconnects, programmable I/Os,
  and  development of new FPGAs and field-configurable memories.

o Novel applications of FPGAs, including rapid prototyping, logic
  emulation, reconfigurable custom computing, and dynamically
  reconfigurable applications.

o New CAD algorithms and tools for FPGAs,   including  new  algo-
  rithms for sequential and combinational logic  optimization,
  technology mapping, partitioning, placement, routing, and de-
  velopment of new FPGA synthesis or layout systems.

o Advances in field-programmable technology, including  new  pro-
  cess and fabrication technologies, and  field-programmable
  analog arrays.

Authors of papers accepted for the 1997  ACM/SIGDA  5th  Interna-
tional  Symposium on Field-Programmable Gate Arrays (FPGA'97) are
especially encouraged to submit an extended version of their  pa-
pers for possible inclusion in the special issue.  Submissions of
relevant work not presented at FPGA'97 are  also  welcome.   This
special issue is being coordinated by Guest Editors Jason Cong of
Univ. of California, Los Angeles (UCLA) and Carl Ebeling of Univ.
of Washington.

Prospective authors should submit postscript  versions  of  their
papers electronically using the www no later than March 31, 1997,
and indicate they should be considered for this special issue:

         http://microsys6.engr.utk.edu/~tvlsi

Alternatively, email may be used by contacting:

         tvlsi@microsys6.engr.utk.edu

If neither of the electronic means is available to authors, trad-
itional paper manuscripts should be sent to:

     IEEE Trans. on VLSI Systems
         c/o Prof. Bing Sheu
         Electrical Engineering
         Powell Hall, Room 604
         University of Southern California
         Los Angeles, CA 90089-0271

==================================================================
+-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-+
|               Scott A. Hauck, Assistant Professor                         |
|  Dept. of ECE                        Voice: (847) 467-1849                |
|  Northwestern University             FAX: (847) 467-4144                  |
|  2145 Sheridan Road                  Email: hauck@ece.nwu.edu             |
|  Evanston, IL  60208                 WWW: http://www.ece.nwu.edu/~hauck   |
+-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-+
Article: 5040
Subject: Re: ASIC/FPGA Synthesis for LINUX... It's HERE!
From: stefan_doll@usa.racal.com (Stefan Doll)
Date: Wed, 15 Jan 1997 17:31:11 GMT
Links: << >>  << T >>  << A >>
Uwe Bonnes <bon@hertz.ikp.physik.th-darmstadt.de> wrote:

>Has Leonardo some reference page? Some mail adress?

http://www.exemplar.com/leonardo060396.phtml

-Stefan


Article: 5041
Subject: Market Share Stats for Synthesis Vendors?
From: Lance Gin <c43lyg@dso.hac.com>
Date: Wed, 15 Jan 1997 12:42:58 -0800
Links: << >>  << T >>  << A >>
Can anyone share some market share stats for the major ASIC/FPGA 
synthesis vendors (eg. Synopsys, Exemplar, Cadence, Mentor, etc.)?

I'd be interested in either revenue or seat-based statistics. Separate  
stats for the ASIC and FPGA markets would be nice too. Thanks in advance.

-- 
_______________________________________________________________________

Lance Gin                           "off the keyboard, over the bridge,
Delco Systems-GM Hughes Electronics  through the gateway,
C43LYG@dso.hac.com                   nothing but NET!"
_______________________________________________________________________
Article: 5042
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: Henry Spencer <henry@zoo.toronto.edu>
Date: Thu, 16 Jan 1997 04:04:32 GMT
Links: << >>  << T >>  << A >>
In article <32DCEAB8.6C1C@ssec.honeywell.com> David Erstad <erstad@ssec.honeywell.com> writes:
>> Actually, if memory serves, the first requirement isn't quite as stringent
>> as you might think:  it is typically possible to power-cycle DRAM briefly
>> to break a latchup without losing the memory contents...
>
>I think most people and designing and flying space systems would be very 
>uncomfortable with the concept that a part would latch up and it would
>be "OK".

Then they should brace themselves for some uncomfortable news:  the new
generation of designers are already flying systems which cope with latchup
when it happens rather than jumping through hoops on parts selection to
ensure it never does.  For example, the Mars Observer Camera, launched in
1992 (and now en route to Mars, again, under a slightly changed name, on
Mars Global Surveyor), uses non-latchup-proof DRAMs in its main memory.
(In fact, the observation about power cycling without content loss comes
from the MOC people.)
-- 
"We don't care.  We don't have to.  You'll buy     |       Henry Spencer
whatever we ship, so why bother?  We're Microsoft."|   henry@zoo.toronto.edu
Article: 5043
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: ees1ht@ee.surrey.ac.uk (Hans Tiggeler)
Date: 16 Jan 1997 09:05:14 GMT
Links: << >>  << T >>  << A >>
In article <01bc0161$d44246a0$6e0db780@Rich>, rich.katz@gsfc.nasa.gov says...
>
>In the near term, there will be an increasing number of NASA missions
>flying DRAMs, including some long-term missions like Cassini to Saturn.  
>
>The easiest codes to work with are the Hamming codes, with most
>implementations correcting single-bit errors and detecting double-bit
>errors.  These codes permit easy correction on the fly and are very simple
>to implement (and if one has time it can be done if FPGAs).  
>If not, then Reed-Solomon or some other code would be
>required, which are more difficult to decode.  

Given the price of S/DRAM memories why not use a majority voting system such as TMR (Triple 
Modular Redundancy). The system can handle multiple bit upsets and is very easy to implement. I 
have currently a system running in an Actel1020 (17% used!). Of course the drawback is that you 
have to either triplicate your memory or add a statemachine to perform three memory read cycles 
per CPU read cycle.

Hans
Surrey Satellite Technology.

 

Article: 5044
Subject: Re: Safety Critical Apps -> Xilinx Checker.
From: ees1ht@ee.surrey.ac.uk (Hans Tiggeler)
Date: 16 Jan 1997 09:18:02 GMT
Links: << >>  << T >>  << A >>
In article <32D3A689.70E9@item.uni-bremen.de>, beckmann@item.uni-bremen.de 
says...
>
>As far as I could find out, there is no rad hard Xilinx Device
>available.
>
>The application of this was a satellite. 
>
>Fritz
> 
>* Friedrich Beckmann                                          *
>* University of Bremen                  TEL: +49 421 218 4079 *
>* FB1 - Institute for Microelectronics  FAX: +49 421 218 4434 *

Just out of interest has anybody implemented a continuous checking watchdog on 
a Xilinx FPGA? 

Xilinx provides a configuration read back option, using the same device or 
some external CRC checker the system can continuously be checked for 
configuration errors due to SEU. The error latency will of course depends on 
the configuration size and read back speed (1 or 8MHz?). To improve on this 
the user can for example add redundant logic and boundary scan techniques. 
Total dose tolerance can be improved by different packaging techniques such as 
RADPACK (tm) and SEL can be addressed with a fast electronic fuse. 

The unknown of course in this system is, can you damage a Xilinx device by 
randomly changing configuration bits?

Hans.
Surrey Satellite Technology

Article: 5045
Subject: Re: Market Share Stats for Synthesis Vendors?
From: Anatoli Ivanov <anatoli@cadence.msk.su>
Date: Thu, 16 Jan 1997 14:50:11 +0300
Links: << >>  << T >>  << A >>
Lance Gin wrote:
> 
> Can anyone share some market share stats for the major ASIC/FPGA
> synthesis vendors (eg. Synopsys, Exemplar, Cadence, Mentor, etc.)?
> 
> I'd be interested in either revenue or seat-based statistics. Separate
> stats for the ASIC and FPGA markets would be nice too. Thanks in advance.> 

 Try this: http://quote.pathfinder.com/money/quote/qc?symbols=CDN

Best regards. Anatoli

============================================
Anatoli A. Ivanov, Application Engineer

e-mail : anatoli@cadence.msk.su
============================================
Article: 5046
Subject: ILP formulation of DRL Scheduling Problem
From: Peter Sels <peter@easics.be>
Date: Thu, 16 Jan 1997 13:28:49 +0100
Links: << >>  << T >>  << A >>
Hello,

If anybody is interested in an Integer Linear Programming formulation
of the Scheduling Problem for Dynamically reconfigurable FPGAs,
have a look at http://users.ox.ac.uk/~kebl0341/thesis.ps

The idea is that scheduling shouldn't be done manually for FPGAs like:
-XC6000 series of Xilinx
-Pilkington series (licensed by Toshiba and Motorolla)
-Atmel DR FPGAs
because that can be too complex and will lead to non optimal results
that may not meet the design speed criteria.

An algorithm is set up in this thesis, using ILP. In fact just the ILP
formulation is given. I have no time to write a paper. You might want to
write it yourself!

Please tell me what you think about it. I'd be glad to get some
comments.

Peter
-- 
===================================================================
Peter Sels                ===              Easics               ===
ASIC Designer             ===  VHDL-based ASIC design services  ===
mailto:peter@easics.be       ===================================
Tel: +32-16-298 405         Kapeldreef 60, B-3001 Leuven, BELGIUM
Fax: +32-16-298 319                http://www.easics.com
personal homepage: http://users.ox.ac.uk/~kebl0341/
Article: 5047
Subject: Results: Linux version for EDA
From: "Karl W. Pfalzer" <karlp@ez-synthesis.com>
Date: Thu, 16 Jan 1997 08:59:46 -0800
Links: << >>  << T >>  << A >>
Last week, I posted a survey to ascertain the version of Linux
users have (or intend to use) for EDA tools.

I received 47 responses; 44 of which contributed survey data
(the other 3 just offered encouragement to support Linux for EDA usage).

The results are:

 8.5% use Linux 1.2
85.1% use Linux 2.0
 6.4% use Linux 2.1

The additional query about SunOS yielded the following distribution
for those who responded that they use Sun machines:

47.6% use SunOS 4.1
 9.5% use SunOS 5.4 (Solaris 2.4)
42.9% use SunOS 5.5 (Solaris 2.5)

Much thanx to those who responded.

BTW, our company is very interested in developing the "right" tools for
Linux (et. al.) 
platforms.  Initially, the focus will be on tools to complement a
Synopsys Design Compiler
based synthesis flow:

  1.  GUI-based front end to Design Compiler which offers shrink-wrapped
capability 
      to perform true, easily controlled, hierarchical ASIC synthesis
with effective
      budgeting techniques --- without writing one line of dc_shell
code.
  
      (The tool also generates "scripts" which can be leveraged,
replayed w/o GUI !)

  2.  Write_script "preprocessor" which offers a very easy to use Tcl
      (database/procedural) interface to massage
characterize/write_script constraints 
      before subsequent use.  A useful set of predefined routines are
provided
      (to coalesce bit/bus constraints, select AC/DC categories, ...). 
    
  3.  Tcl (database/procedural) interface to a (post-synthesis) netlist
to provide
      additional (directed) manipulations/optimizations (an infinitely
"programmable"
      change-names, buffer insertion/removal, sizing, ...).  Several
predefined routines
      will be provided.

The company's (current) intent is to utilize the Internet as the sole
means of
distribution and support: i.e., the virtual EDA company.

Obviously, ease-of-use, quality documention, timely FAQs are important
to being
successful with this model.
 
Any comments, suggestions, etc. concerning these tools is welcome.

(Personal email may be more appropriate than posting to the newsgroups.
Summary responses may be posted, subsequently, unless massive "flames"
result.)

Thanx.

        
-- 
Synthesis Solutions, Inc.
http://www.ez-synthesis.com
vmail: (415) 431-6429
Article: 5048
Subject: Re: ASICs Vs. FPGA in Safety Critical Apps.
From: "Rich K." <rich.katz@gsfc.nasa.gov>
Date: 16 Jan 1997 17:45:57 GMT
Links: << >>  << T >>  << A >>
designs of this type have been done before and we called it a poor man's
edac (and it was done when they had 1 kbit memories <- not a typo).  didn't
even need an fpga, the 54ls253 makes a great voter w/ only one half and by
adding an inverter the other half can function as an error indicator.  and
the state machine wasn't tough and the read cycle had to be transformed
into a read-modify-write; this was important for consistent software
timing/verification in the presence of errors.

rk

p.s. and i did a 16-bit flow through edac in a 1020 which had the ability
(when using x8 ram chips) to swap out failed chips in the real memory and
use the spare edac memory for functioning; handy for when a whole chip gets
busted, hamming codes won't fix that problem, but gotta keep going and
don't have the room for 3 copies of memory.

Hans Tiggeler <ees1ht@ee.surrey.ac.uk> wrote in article
<5bkr0a$ru4@info-server.surrey.ac.uk>...
> In article <01bc0161$d44246a0$6e0db780@Rich>, rich.katz@gsfc.nasa.gov
says...
> >
> >In the near term, there will be an increasing number of NASA missions
> >flying DRAMs, including some long-term missions like Cassini to Saturn. 

> >
> >The easiest codes to work with are the Hamming codes, with most
> >implementations correcting single-bit errors and detecting double-bit
> >errors.  These codes permit easy correction on the fly and are very
simple
> >to implement (and if one has time it can be done if FPGAs).  
> >If not, then Reed-Solomon or some other code would be
> >required, which are more difficult to decode.  
> 
> Given the price of S/DRAM memories why not use a majority voting system
such as TMR (Triple 
> Modular Redundancy). The system can handle multiple bit upsets and is
very easy to implement. I 
> have currently a system running in an Actel1020 (17% used!). Of course
the drawback is that you 
> have to either triplicate your memory or add a statemachine to perform
three memory read cycles 
> per CPU read cycle.
> 
> Hans
> Surrey Satellite Technology.
> 
>  
> 
> 
Article: 5049
Subject: Re: Safety Critical Apps -> Xilinx Checker.
From: "Rich K." <rich.katz@gsfc.nasa.gov>
Date: 16 Jan 1997 18:11:36 GMT
Links: << >>  << T >>  << A >>
well the xilinx guys just left the office and the amount of damage done
internally by an illegal configuration they didn't know but i'll be
following up at the factory with the contact number they left.  a key
element would be worst-case current densities so there is no metal
migration problems.  the standard for this is about 2E5 a/cm^2, if memory
is correct, which is not as big as it looks when you compute the
cross-sectional area of small metal lines.

another case for damage would be that of an input cell turning into an
output with a single bit-flip and that is obviously a concern.

as far as continously checking the configuration, we discussed that and
hope to test some of their devices fabricated on epi (to hopefully make
latchup problems go away) and felt we could do a verify cycle in < 10 ms. 
it is planned to implement this and test it in an accelerator soon.  of
course you need a master copy of the program and where do you store that? 
and a sequencer to take care of things if there is a failure to verify. 
for the test i'd like to run, i'll just use a eprom and an actel a1020 :-)
to do the work.

as for the 'fast electronic fuse,' i'd be interested in seeing a good
design for hi-rel apps that is testable. it's talked about a lot but
difficult and bulky to do well.

additionally, with any total dose radiation shielding, practical shielding
materials/thickness/mass/configurations often limit the benefits of
shielding, depending upon the environment.  electrons are relatively easily
to shield.  protons are much more difficult and very soft commercial
devices, even with thick shields, may not meet mission requirements.  of
interest is that shielding can even make matters worse for some particles
depending upon the energy.  it has been showed analytically that the shield
will change the energy of particles that are not stopped by the shield
(difficult to do for protons, for instance) with the lower energy causing
more damage to the device.  this was recently measured for several
different types/thicknesses of shields using soft fpga's and for certain
ranges of energies, the "shielded" parts were damaged faster than
unshielded parts.

rk

Hans Tiggeler <ees1ht@ee.surrey.ac.uk> wrote in article
<5bkroa$ru4@info-server.surrey.ac.uk>...
> In article <32D3A689.70E9@item.uni-bremen.de>,
beckmann@item.uni-bremen.de 
> says...
> >
> >As far as I could find out, there is no rad hard Xilinx Device
> >available.
> >
> >The application of this was a satellite. 
> >
> >Fritz
> > 
> >* Friedrich Beckmann                                          *
> >* University of Bremen                  TEL: +49 421 218 4079 *
> >* FB1 - Institute for Microelectronics  FAX: +49 421 218 4434 *
> 
> Just out of interest has anybody implemented a continuous checking
watchdog on 
> a Xilinx FPGA? 
> 
> Xilinx provides a configuration read back option, using the same device
or 
> some external CRC checker the system can continuously be checked for 
> configuration errors due to SEU. The error latency will of course depends
on 
> the configuration size and read back speed (1 or 8MHz?). To improve on
this 
> the user can for example add redundant logic and boundary scan
techniques. 
> Total dose tolerance can be improved by different packaging techniques
such as 
> RADPACK (tm) and SEL can be addressed with a fast electronic fuse. 
> 
> The unknown of course in this system is, can you damage a Xilinx device
by 
> randomly changing configuration bits?
> 
> Hans.
> Surrey Satellite Technology
> 
> 


Site Home   Archive Home   FAQ Home   How to search the Archive   How to Navigate the Archive   
Compare FPGA features and resources   

Threads starting:
1994JulAugSepOctNovDec1994
1995JanFebMarAprMayJunJulAugSepOctNovDec1995
1996JanFebMarAprMayJunJulAugSepOctNovDec1996
1997JanFebMarAprMayJunJulAugSepOctNovDec1997
1998JanFebMarAprMayJunJulAugSepOctNovDec1998
1999JanFebMarAprMayJunJulAugSepOctNovDec1999
2000JanFebMarAprMayJunJulAugSepOctNovDec2000
2001JanFebMarAprMayJunJulAugSepOctNovDec2001
2002JanFebMarAprMayJunJulAugSepOctNovDec2002
2003JanFebMarAprMayJunJulAugSepOctNovDec2003
2004JanFebMarAprMayJunJulAugSepOctNovDec2004
2005JanFebMarAprMayJunJulAugSepOctNovDec2005
2006JanFebMarAprMayJunJulAugSepOctNovDec2006
2007JanFebMarAprMayJunJulAugSepOctNovDec2007
2008JanFebMarAprMayJunJulAugSepOctNovDec2008
2009JanFebMarAprMayJunJulAugSepOctNovDec2009
2010JanFebMarAprMayJunJulAugSepOctNovDec2010
2011JanFebMarAprMayJunJulAugSepOctNovDec2011
2012JanFebMarAprMayJunJulAugSepOctNovDec2012
2013JanFebMarAprMayJunJulAugSepOctNovDec2013
2014JanFebMarAprMayJunJulAugSepOctNovDec2014
2015JanFebMarAprMayJunJulAugSepOctNovDec2015
2016JanFebMarAprMayJunJulAugSepOctNovDec2016
2017JanFebMarAprMayJunJulAugSepOctNovDec2017
2018JanFebMarAprMayJunJulAugSepOctNovDec2018
2019JanFebMarAprMayJunJulAugSepOctNovDec2019
2020JanFebMarAprMay2020

Authors:A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Custom Search